AI Agents Drive Cyber Attacks Higher, Raising 2-Sided Escalation Risks
Updated
Updated · Financial Times · Oct 8
AI Agents Drive Cyber Attacks Higher, Raising 2-Sided Escalation Risks
3 articles · Updated · Financial Times · Oct 8
Summary
Australian officials last month disclosed an AI-agent breach of healthcare systems, underscoring how criminal and state-backed actors are already using AI in live intrusions.
Anthropic said two weeks earlier that Russian cyber actors were using AI for reconnaissance, hacking and data theft more cheaply and efficiently, accelerating the shift toward more autonomous operations.
That autonomy clouds attribution for defenders and weakens control for attackers, because AI can adapt inside networks in ways humans did not explicitly direct or may struggle to predict.
Russian-style deception tactics such as maskirovka make that ambiguity more dangerous, since defenders may read machine-driven lateral movement as sabotage preparation rather than espionage.
The result is a deterrence problem on both sides: actors gain capability by delegating decisions to AI, but also increase the risk of miscalculation and unintended escalation.